ODIN AI Task Planner module
Main functionalities
The AI task planner module of ODIN focuses on the online and offline task planning process for a selected assembly operation. ODIN task planner is able to be triggered during assembly operation’s execution in case of unexpected events. The task planner tool is able to automatically analyse and assess the production tasks and rearrange the available production resources/modules in real-time.
Firstly, a set of assembly tasks are inserted in the AI task planner module by the production manager in order to for scenario’s modelling. Then, the scenario is added in the ODIN world model including the resource and workload hierarchical modelling of the pilot. The search engine of ODIN AI task planner module generates different alternative task plans based on tasks’ model and resources’ capabilities.
These task plans are stored in the ODIN database for validation and evaluation through the ODIN Digital Simulation. The Digital Simulation module is connected with the ODIN AI task planner module for the simulated validation of each generated task plan. Information about the simulated execution of each task plan are stored in ODIN database. Digital Simulation layout initialization is based on data exported from the Digital Twin and the OpenFlow modules of ODIN. In order to avoid multiple execution of the same task by a specific resource when the initial conditions for the resource are the same, ODIN AI task planner module will check if task’s evaluation data are already available in the database.
After the simulated execution of each task plan, the AI based task planner module is responsible to retrieve data about each task plans’ execution from the database and evaluate them. Based on the evaluation criterias’ weight defined by the production manager, the best alternative task plans will be visualized through the UI to the production manager. He/She is able to evaluate them based on his/her knowledge and select the best alternative to be sent for physical execution through the OpenFlow.
Technical specifications
ODIN task planning algorithm consists of 4 steps:
- Receive request for Task Planning: The task planner algorithm receives a request for execution either by the production manager or dynamically by the OpenFlow in case of unexpected events in the shopfloor forcing the production to stop.
- Initiate and trigger Task Planner through the developed UI: The production manager initializes the search parameters, the evaluation parameters’ weights, the filtering criteria and the number of plans to be calculated for the task planning algorithm. Then the task planner algorithm is triggered to start calculating task plans and sends the generated task plans for validation in the Digital Simulation.
- Task Plan validation and evaluation: In this step, the AI task planner module’s connection with the Digital Simulation but also the Digital Twin is presented. The generated task plans from the Task Planner algorithm are sent for simulated execution in the Digital Simulation of ODIN. Digital environment’s layout initialization is based on data received from the OpenFlow (robot’s joint values, fixed resources’ location etc.) and the Digital Twin (operator’s location). OpenFlow is responsible to receive the tasks’ allocation from the AI task planner module and translate them to primitive actions for simulated execution in the 3D simulation. The Digital Simulation executes the primitive actions and stores information about these actions’ execution in ODIN database to be used by the OpenFlow during task plans’ evaluation.
- Generated Task Plan evaluation and best alternative selection for execution: The generated task plans are automatically evaluated based on the pre-defined evaluation metrics and their selected weights. Then the top 3 task plans are visualized through this modules’ UI. The production manager is able to check the AI task planner module’s result through the UI and save the optimal task plan for execution in the physical environment.
Owner (organization)
ODIN Project Coordinator
Dr. Sotiris Makris
Laboratory for Manufacturing Systems and Automation (LMS),
University of Patras
Tel: +30-2610-910160
Fax: + +30-2610-997314
e-Mail: makris@lms.mech.upatras.gr